NCASEAI / Research guides
AI market research for Saudi Arabia that cannot invent a number
AI is quick at research and prone to making numbers up. NCASEAI keeps the speed and removes that risk: an AI model writes the sentences, code inserts every official figure from a sourced ledger, and a second model from a different company checks the result.
No credit card required · Arabic & English
What you can ask
- Who are the main AI companies in Saudi Arabia?
- What is the maximum penalty under the PDPL?
- How is the Saudi government investing in AI?
- Summarise the Saudi AI market for a board paper.
Example questions. NCASEAI answers each from its sources, or names the gap.
The four parts
- A language model writes. It drafts the answer and, for a report, its sections. It is never allowed to type an official figure: a draft with a number it was not given is refused.
- Retrieval finds the evidence. Official figures come from a ledger where every row has a publisher, period and date; web pages come from a live search, each kept with its address.
- A fixed workflow runs the steps: plan, retrieve, write, insert, check, show. Code runs it, so no step is skipped, and the steps stream live while it works.
- Source verification closes it. Code inserts each figure from its ledger row, and a second model checks every statement and every web figure against its source. What fails is marked.
What it reads today
The Saudi publishers NCASEAI reads today. A source that is not connected is never guessed at.
- Connected
- ZATCA, SAMA, World Bank
- Partly connected
- GASTAT (selected tables, through DataSaudi, as Other source)
- Not connected yet
- MISA, Etimad, CST, SFDA
AI works on top of the sources below. It cannot add a source that is not connected.
What you get
- An answer written with AI and labelled as such
- A visible check by a second AI model from a different company
- A report you can edit, then export to PDF, Word or PowerPoint
- A plain list of the models used and what each one does, in the Trust Center
Questions
Which AI models does NCASEAI use?
A model from Anthropic writes and a model from OpenAI checks, both through OpenRouter, with Exa for web search. The Trust Center lists the current defaults.
Is my data used to train AI models?
No. NCASEAI does not use your content to train models, and every model request asks to be routed only to providers with zero-data-retention terms.
Can the AI still be wrong?
Yes. The check can show that a figure is not in its source; it cannot tell whether the source is right. Open the sources before you rely on a figure from the web.